Design And Application Of Robust Stability Guaranteed Disturbance Observer For Multi-variable Systems

The ability to accomplish a task is highly depended on its high accuracyof motion control. Feedback control only cannot fulfill the need of controlperformance. This growing demand for high accuracy control has stimulateddevelopments of advanced control methods and explorations of new controltheories. External disturbances and modeling errors are the primary factorsthat deteriorate the performance of control systems. So far, several attemptshave been made to achieve precise performance in the presence of both ex-ternal disturbances and modeling errors. Among these, disturbance observer(DOB) has been widely employed in industrial applications due to its suc-cinct structure, ability to reject disturbances, and robustness against modelperturbations.However, the implementation of DOB on multi-variable plants, com-paring to those on single-input single-output (SISO) plants, is extremely lim-ited. The two main reasons are: a) the conventional DOB is designed specif-ically for SISO plants; b) it is difficult to stabilize the DOB-based mul-ti-variable system against uncertainties. In this thesis, we mainly discuss the implementation of disturbance ob-server in multi-variable systems, especially the guarantee of closed-loop ro-bust stability. Our main works include:(1) Constructed a proper structure of multi-variable disturbance observ-er control system, which includes as much information as possible in the nominal plant, while sacrificing the multi-variable structure for simplicity in the design of Q-filter.(2) Analyzed the closed-loop robust stability, disturbance rejection abil-ity and the influence of sensor noise of linear multi-variable disturbance ob-server system. Thereby systematically designed the Q-filter with optimum disturbance rejection ability and guaranteed closed-loop robust stability bas-ing on H∞robust control theory.(3) Similar analysis has been done on nonlinear multi-variable systems and systematic design of Q-filter was reached.(4) Simulations and experiments on a quadrotor UAV demonstrated the efficacy of the proposed method.